Method for Converting Speech Signal to Improve Speech Intelligibility

The task of improving the speech intelligibility in communication systems is considered. The acute problem of speaker voice recognition when using known methods for solving it is pointed out. To overcome this problem, new method for converting a speech signal is proposed. It based on an autoregressive model of the vocal tract and on the principle of frequency-selective amplification of the main formants. An example of the practical implementation of a new method based on the fast Fourier transform is considered. Estimates of computational costs and its performance are given. A full-scale experiment was set up and carried out. Based on its results, the positive effect achieved by applying the proposed method was established, namely: increasing the intelligibility of the speech of the control speaker while maintaining a sufficiently high degree of recognition of his voice. The results obtained are intended for use in the development of new and modernization of existing voice communication systems, including a mobile communication and VoIP-systems.